Department of Nutritional Sciences, University of Wisconsin-Madison, Madison, Wisconsin, USA
Waisman Center for Neurodevelopmental Disorders, University of Wisconsin-Madison, Madison, Wisconsin, USA
Address correspondence and reprint requests to Susan Smith, Department of Nutritional Sciences, University of Wisconsin-Madison, 1415 Linden Dr, Madison WI 53706, USA. E-mail: email@example.com
Prenatal ethanol exposure causes persistent neurodevelopmental deficits by inducing apoptosis within neuronal progenitors including the neural crest. The cellular signaling events underlying this apoptosis are unclear. Using an established chick embryo model, we previously identified ethanol's activation of calmodulin-dependent protein kinase II (CaMKII) as a crucial early step in this pathway. Here, we report that CaMKII is pro-apoptotic because it mediates the loss of transcriptionally active β-catenin, which normally provides trophic support to these cells. β-catenin over-expression normalized cell survival in ethanol's presence. CaMKII inhibition similarly restored β-catenin content and transcriptional activity within ethanol-treated cells and prevented their cell death. In contrast, inhibition of alternative effectors known to destabilize β-catenin, including glycogen synthase kinase-3β, Protein Kinase C, JNK, and calpain, failed to normalize cell survival and β-catenin activity in ethanol's presence. Importantly, we found that purified CaMKII can directly phosphorylate β-catenin. Using targeted mutagenesis we identified CaMKII phosphorylation sites within human β-catenin at T332, T472, and S552. This is the first demonstration that β-catenin is a phosphorylation target of CaMKII and represents a novel mechanism by which calcium signals could regulate β-catenin-dependent transcription. These results inform ethanol's neurotoxicity and offer unexpected insights into other neurodevelopmental and neurodegenerative disorders having dysregulated calcium or β-catenin signaling.
We proposed that ethanol initiates apoptosis in early neuroprogenitors through an intracellular calcium transient that activates CaMKII, which in turn destabilizes transcriptionally-active beta-catenin, perhaps through direct phosphorylation at Thr332, Thr472 and Ser552. This represents a novel mechanism of calcium-modulation of Wnt signaling and of ethanol's neurotoxicity.
Fetal alcohol spectrum disorders (FASD) are a leading cause of neurodevelopmental disability in most western societies. The most severe form, fetal alcohol syndrome, affects 0.33 to 2.2 per 1000 live births, with substantially higher rates (8 to 89.2 per 1000) in populations with appreciable alcohol use (May et al. 2009). Prenatal alcohol exposure (PAE) causes significant apoptosis within developing neuronal populations, and these losses contribute to the pathologies underlying FASD (Cartwright et al. 1998; Ikonomidou et al. 2000; Dunty et al. 2001). An early neural progenitor population whose development is disrupted by ethanol is the neural crest, a unique stem cell population that originates within the dorsal neuroectoderm (Le Douarin and Kalcheim 1999). As the elevated neural folds fuse, neural crest cells emigrate into peripheral tissues and contribute to diverse structures, including the sympathetic and parasympathetic nervous system, elements of the cranial ganglia, and craniofacial bone and cartilage. Clinically relevant ethanol concentrations (0.05%–0.3%, or 8.6–52 mM) cause substantial apoptosis within neural crest progenitors as revealed by the cells’ pyknotic appearance, their detection using terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL), and dependence of this cell death upon caspase activity (Cartwright et al. 1998; Dunty et al. 2001). These neural crest losses contribute to the deficits in cranial nerves, facial structure, and other neurocristopathies that partly typify those with FASD.
To elucidate the mechanism that underlies this ethanol-mediated apoptosis, we investigated the early cellular signaling events that are altered by acute ethanol exposure. Using a chick embryo FASD model, we found that clinically relevant ethanol concentrations (EC50 = 52 mM) cause a rapid (1–2 s), 3-fold elevation of intracellular calcium (Cai+2) stores within the pre-migratory neural crest progenitors that reside in the dorsal neural folds (Debelak-Kragtorp et al. 2003; Garic-Stankovic et al. 2006). We showed that this Cai+2 transient originates from ethanol's stimulation of G-protein coupled signaling via Gαi2/3 and Gβγ, the latter of which activates phospholipase Cβ and phosphoinositide-mediated release of Cai+2 stores (Garic-Stankovic et al. 2005, 2006). Pre-treatment with Cai+2 chelators prevents the ethanol-induced Cai+2 rise and subsequent cell death, thus normalizing neural crest survival (Debelak-Kragtorp et al. 2003; Garic-Stankovic et al. 2005). How this ethanol-induced Cai+2 transient initiates neural crest apoptosis is unclear. However, we recently identified two signals downstream of the Cai+2 transient that are essential participants in ethanol's pro-death signal. The first is CaMKII, whose activity in pre-migratory neural crest increases 3-fold within 1 min of ethanol addition (Garic et al. 2011). We found that forced expression of CaMKII is sufficient to cause neural crest apoptosis in ethanol's absence, whereas inhibition of CaMKII activity prevents this cell death and rescues neural crest survival in ethanol's presence (Garic et al. 2011). The second key signal is the transcriptional effector β-catenin, which is essential for neural crest induction, migration, and survival (Stuhlmiller and Garcia-Castro 2012). We found that nuclear β-catenin is selectively depleted from neural crest within 2 h of ethanol addition, and this loss is mediated by ethanol's Cai+2 transient (Flentke et al. 2011). Forced expression of β-catenin similarly protects neural crest from ethanol-induced cell death and rescues their survival (Flentke et al. 2011). This loss of β-catenin protein is post-transcriptional, as its transcript levels are unaffected by ethanol.
It is unclear whether ethanol's effects upon CaMKII and β-catenin are mechanistically linked. However, studies in other models suggest a potential interaction between them. Two intersecting pathways act post-translationally to regulate β-catenin stability and thereby its transcriptional activity. In the Canonical Wnt pathway, secreted Wnt glycoproteins interact with their cell-surface Frizzled receptors to inhibit the protein kinase GSK3β (reviewed in MacDonald et al. 2009). In the presence of canonical Wnt/Frizzled signals, β-catenin interacts with T cell transcription factor/lymphoid enhancer factor (TCF/LEF) proteins to initiate transcription. When Wnt signals are absent, GSK3β is no longer silenced and it phosphorylates β-catenin, targeting β-catenin for ubiquitination and proteosomal destruction. In the second pathway, the Non-canonical Ca/Wnt pathway, Wnt interactions with G proteins elevate Cai+2 and repress β-catenin's transcriptional activity through both transcriptional and post-transcriptional means (reviewed in Kohn and Moon 2005). This pathway is GSK3β-independent and utilizes several calcium effectors in a cell-dependent manner. CaMKII can indirectly silence β-catenin-mediated transcription through its phosphorylation of both LEF-1 and the TGF-ß-activated kinase/Nemo-like kinase (TAK/NLK) kinases that negatively modulate TCF/LEF activity (Ishitani et al. 2003). Protein Kinase C can antagonize canonical Wnt signaling through its phosphorylation of Wnt pathway effectors, including dishevelled and β-catenin itself (Gwak et al. 2006). Calcineurin negatively regulates β-catenin via its nuclear factor of activated T cells targets to decrease β-catenin transcription (Saneyoshi et al. 2002). Finally, the calcium-activated proteases, calpains, can directly cleave β-catenin and inactivate the protein (Li and Iyengar 2002). We also note that ethanol itself enhances GSKβ3 activity in selected neuronal populations (French and Heberlein 2009; Liu et al. 2009), although the consequences of this to β-catenin have yet to be examined.
Here, we evaluate how these calcium-dependent and -independent effectors of β-catenin contribute to ethanol's neurotoxicity in the neural crest. We show that activated CaMKII, and not other Wnt effectors, selectively mediates both the loss of transcriptionally active β-catenin from these cells and their subsequent death. β-catenin gain-of-function and CaMKII inhibition normalizes canonical Wnt signaling within these cells and rescues them from ethanol's neurotoxicity. Importantly, we show that purified CaMKII can directly phosphorylate β-catenin in a cell-free context, and we identify those target sites within the protein. This work uncovers a novel mechanism by which calcium signals could regulate the transcriptional activity of the β-catenin/TCF-LEF complex.
Embryos and ethanol treatment
Fertile White Leghorn chick eggs (W98, Hyline, Spencer, IA, USA; Special Black, Sunnyside Farms, Beaver Dam, WI, USA) were incubated to the desired developmental stage (Hamburger and Hamilton 1951). For in ovo studies, control saline or 0.43 mmol ethanol in isotonic saline was injected into the egg yolk center at stage 8/9 (3–7 somites); this produces a peak embryonic ethanol concentration of 50–60 mM for 1–1.5 h post-injection (Debelak and Smith 2000). Ex vivo studies incubated stage 8/9 embryos 2 h in Tyrode's buffer ± 52 mM ethanol; this ethanol concentration causes half-maximal calcium release in these embryos (Garic-Stankovic et al. 2006). All embryos were precisely matched in somite number.
Prior to ethanol challenge some embryos were pre-treated with selective pharmacological compounds to investigate signaling pathway contributions. Membrane-permeable forms were used and included N-Acetyl-Leu-Leu-Nle-CHO (ALLN) (20 μM), Bapta-AM (1 mM), 1-methyl-BIO (1 μM), calmidazolium (10 μM), Calpain Inhibitor V (2 mM), Gö-6983 (50 μM), JNK Inhibitor II (1 μM), Ro-32-0432 (10 μM; all from Calbiochem, San Diego, CA, USA), myristylated-CaMKII Autoinhibitory Peptide (myr-AIP, 2.5 μM, BioMol Research Laboratories, Plymouth Meeting, PA, USA), ionomycin (50 μM, Sigma, St Louis, MO, USA), and SB-415,286 (1 μM, Sigma). For in ovo treatment, compounds were transiently delivered using hydrophobic (SM2, 100 μm diameter) or anion-exchange beads (AG-50W, 75 μm, both from Bio-Rad, Hercules, CA, USA) that were presorbed with the agent and washed prior to implant. Pilot studies ascertained the appropriate concentration to be tested; because bead-mediated delivery is diffusion dependent, the bead soaking concentrations were generally 100- to 1000-fold greater than levels administered by direct delivery (Garic-Stankovic et al. 2005; Garic et al. 2011). Beads were placed adjacent the Hamburger-Hamilton stage (HH8) presumptive hindbrain. 3 h later embryos were treated with saline or ethanol, and beads were removed by gentle aspiration 3 h thereafter. Controls received beads treated with the dimethylsulfoxide carrier solvent at final concentrations ≤ 0.1%. Cell death was evaluated 20 h later, at HH12/13- (16–18 somites), as described below. For ex vivo treatment, HH8 embryos were incubated in the compound for 15 min prior to saline or ethanol challenge, and their β-catenin protein levels were evaluated 2 h later by immunostain or western blot as described below.
Cell death analysis
Cell death was visualized at HH12/13- using LysoTracker Red (LTR) (0.5 μM, Invitrogen, Carlsbad, CA, USA), which we and others have shown detects apoptotic death in the early embryo FASD model (Cartwright et al. 1998; Dunty et al. 2001). To quantify neural crest cell death, the LTR-stained hindbrain was sectioned and immunostained for the neural crest markers Slug (aka Snail2, #27568, 1/1000; AbCam, Cambridge, MA, USA) or Sox9 (1/1000; Abcam) followed by Alexa488-conjugated secondary antibody. We counted the total number of slug+ (Sox9+) and LTR+slug+ (Sox9+) cells residing in the dorsal neural roof and adjacent lateral quadrant of hindbrain rhomobomeres 4 (r4) and r6, which normally do not experience significant cell death. We enumerated 6–10 sections per embryo and 3–4 embryos per treatment, counting > 100 slug+(Sox9+) cells per embryo and normalized to the number of sections counted. In the small molecule screen, we counted the number of LTR+ cells in r4. The screens evaluated 15–20 embryos per treatment and were performed in at least triplicate.
HH8 embryos were incubated 15 min with the indicated compound, washed, treated 2 h with saline or 52 mM ethanol and then were fixed in Dent's solution, sectioned, and immunostained for β-catenin with antibody #15B8 (1 : 1000, Novus Biologicals, Littleton, CO, USA) and Alexa 488-conjugated secondary antibody. 4′,6-diamidino-2-phenylindole (DAPI) counter-stain visualized nuclei. Photographic exposure time was uniform between images.
This was performed as described (Flentke et al. 2011). Ex vivo somite-matched HH8/9 embryos were pre-treated as above with the indicated compound and then challenged 2 h with saline or 52 mM ethanol. Proteins extracted from dissected crania were identified using antibodies against β-catenin (15B8, 1 : 50 000) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (#G8795, 1 : 50 000; Sigma). Experiments were performed in triplicate and analyzed one dissected head per lane and three to five embryos per treatment.
Electroporation and cell death assays
This was performed as described (Flentke et al. 2011). cDNA encoding enhanced green fluorescent protein (eGFP) and the parent expression vector pCAGGS (Niwa et al. 1991) were kind gifts of T. Suzuki. The β-catenin construct was obtained from Addgene (#13435; Cambridge, MA, USA; Lee et al. 2001). The ΔTCF construct (Addgene #14019; Kardon et al. 2003) lacks the β-catenin binding domain (amino acids 2–50) and thus has dominant-negative activity. Constructs were used at a 3 : 1 ratio (1.2 : 0.4 μg/μL) of experimental:eGFP cDNA. DNA (2 μL) was injected into the lumen of the HH9 posterior hindbrain, and the in ovo embryos were immediately electroporated as described (Flentke et al. 2011). Eggs were injected 3 h later with saline or 0.43 mmol ethanol. At 20 h thereafter (HH12/13-), cell death and neural crest numbers were enumerated as above using LysoTracker Red and Sox9 immunostain. Embryos lacking hindbrain expression of the parent eGFP vector were discarded from analysis.
Quantitation of endogenous β-catenin/TCF transcriptional activity
This was performed as described (Flentke et al. 2011). In ovo embryos were electroporated with the β-catenin/TCF luciferase reporter construct pBARL or the parallel, inactive construct pfuBARL (5 μg/μL; kind gift of R. Moon; Biechele and Moon 2008) + cytomegalovirus (CMV)-Renilla luciferase vector (0.06 μg/μL; Promega, Madison, WI, USA). Prior to electroporation some embryos were treated 3 h with Bapta-AM or calmidazolium as described above. Luciferase activity of individual crania was quantified 20 h thereafter using a modification of the Dual Luciferase Reporter Assay System (Promega) as described (Flentke et al. 2011). Luciferase values were normalized to Renilla activity. Extracts were assayed in duplicate using six to eight embryos per treatment; experiments were repeated in at least duplicate.
In vitro phosphorylation studies
cDNA encoding human recombinant β-catenin in pET28a (#17198, Addgene; Xing et al. 2008) was expressed and purified using a cobalt affinity column (Clontech, Mountain View, CA, USA). The eluate was treated with tobacco etch virus protease (Promega) to cleave the His-tag and repurified as above before use. Site-directed mutagenesis to substitute Alanine at Threonine-332, Serine-389, Threonine-472, and/or Serine-552 was performed using the Quick Change Lightning mutagenesis kit (Stratagene, Cedar Creek, TX, USA) according to manufacturer's directions. Truncated rat recombinant CaMKII (AA 1-325) was prepared in Sf9 baculovirus system and obtained from New England Biolabs (Beverly, MA, USA); it was activated by phosphorylation following manufacturer's guidelines. Unactivated CaMKII controls were run in parallel with omission of calcium and calmodulin. The protein D52, a known CaMKII substrate, was a generous gift of G. Groblewski (Chew et al. 2008). For the reaction, recombinant β-catenin (0.96 μg) or D52 (0.1 μg) was incubated (3 h, 30°C) with either activated (11 units) or unactivated (11 unit equivalents) recombinant CaMKII, in a CaMKII reaction mixture (New England Biolabs) containing 50 mM Tris-HCl, 10 mM MgCl2, 0.1 mM EDTA, 0.01% Brij-35, 0.3 mM ATP, 2.5 μCi [γ32P]-ATP (Perkin-Elmer, Shelton, CT, USA). This represented three molar equivalents of β-catenin (89.9 kD) over D52 (28 kD). Samples were separated using 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis, transferred to Immobilon-P membrane, and autoradioimages were acquired using a Typhoon FLA-7000 (GE BioSciences, Piscataway, NJ, USA).
Normally distributed data were subjected either to unpaired t-test or one-way anova and the appropriate post hoc analysis as indicated, using SigmaStat v.3.5 (Systat Software, Point Richmond, CA, USA). Data not normally distributed were analyzed using Kruskal–Wallis one-way anova on ranks and the non-parametric version of Dunn's post hoc analysis, again using SigmaStat. p < 0.05 was the criterion for significance. Results are presented as mean ± SEM, unless otherwise indicated.
Cai+2 mediated loss of β-catenin and neural crest
As we showed previously (Cartwright et al. 1998), exposure of the neurulation-stage chick embryo (three to five somites) to clinically relevant ethanol concentrations (50–60 mM) initiates substantial cell death within the early neural tube and hindbrain. Many of the affected cells are of neural crest origin, revealed by immunostain for the neural crest marker Sox9+ and their location within the dorsal hindbrain roof (Fig. 1a, b). The level of neural crest death rises from 8.7 ± 4.5% to 35.5 ± 7.6% (p < 0.001) following ethanol exposure (Fig. 1e), and neural crest numbers drop to 60% of normal (p < 0.001, Fig. 1f). As we reported previously (Debelak-Kragtorp et al. 2003; Garic-Stankovic et al. 2005), a well-documented ethanol-invoked Cai+2 transient initiates this cell death. Treatment with the intracellular calcium chelator Bapta-AM immediately prior to ethanol exposure normalized neural crest numbers (35.1 ± 2.5, Fig. 1f) and prevented the ethanol-induced cell death (4.8 ± 6.2%, both p < 0.001 vs. ethanol; Fig. 1c, e). In the absence of ethanol, treatment with the calcium ionophore Ionomycin initiated significant neural crest death (41.3 ± 10.7%, p < 0.001; Fig. 1d, e) to levels indistinguishable from those caused by ethanol, showing that a transient elevation in Cai+2 is sufficient to cause neural crest death.
At these stages (HH8, three to five somites), the dorsal neuroepithelium including neural crest progenitors was selectively enriched in β-catenin protein (Fig. 2a, b), as we and others reported previously (Brault et al. 2001; Flentke et al. 2011). A 2 h exposure to 52 mM ethanol substantially down-regulated β-catenin protein levels within these dorsal neuroepithelial populations, including neural crest progenitors within the dorsal lip of the neural folds (boxed region, Fig. 2a, b). Western blot analysis of ethanol-treated hindbrain confirmed these observations and showed that β-catenin protein content decreased to 56 ± 10% of control values (Fig. 2c, d; p < 0.005). The ethanol-mediated loss of β-catenin was calcium-dependent. Pre-treatment with Bapta-AM prevented the β-catenin loss and normalized its expression in the presence of ethanol (Figs. 2a–d; p = 0.037 vs. ethanol). BAPTA pre-treatment also exaggerated neural fold elevation in six of seven embryos, an action consistent with calcium's known role within the developing neural plate (Ferreira and Hilfer 1993). In contrast, treatment of otherwise-normal embryos with ionomycin caused a significant loss of β-catenin protein as evaluated using immunostain (Fig. 2a, b) and western blot (Fig. 2c, d). The level of loss was similar to that caused by ethanol (52 ± 9%, p = 0.003 vs. saline; p = 0.74 vs. ethanol). This suggested that the β-catenin protein in neural crest progenitors was sensitive to Cai+2 –mediated destabilization and degradation. To further investigate the calcium mechanism, we pre-treated neural folds with the calmodulin inhibitor calmidazolium (Gietzen et al. 1982); calmodulin is a cytoskeletal calcium sensor protein that converts the calcium transient into a lasting effector of cell fate through the calcium-dependent interactions with protein kinases and phosphatases. Calmidazolium pre-treatment prevented the β-catenin loss in ethanol-treated neural crest (p = 0.003 vs. ethanol; Fig. 2c, d) indicating that the calcium transient might act upstream through activation of calmodulin-dependent signaling pathways. The β-catenin levels following ethanol + calmidazolium treatment were significantly greater than that of saline controls (p = 0.01 vs. saline), suggesting that calcium may also have a role in controlling endogenous β-catenin levels within this population.
In the nucleus, β-catenin interaction with the transcriptional effectors TCF/LEF1 converts the latter from transcriptional repressors to activators (MacDonald et al. 2009). To ascertain if this β-catenin loss affected its transcriptional function, we transfected hindbrain neural crest with a luciferase-based transcriptional reporter construct and quantified its activity in the presence or absence of ethanol treatment. The pBARL reporter contains twelve TCF response elements that drive luciferase expression (Biechele and Moon 2008). Ethanol challenge significantly reduced the luciferase expression in pBARL-transfected cells as compared with saline-treated controls (41.8 ± 18.0%, p < 0.003, Fig. 2e), indicating that ethanol treatment reduced the transcriptional activity of the β-catenin/TCF complex. Ethanol does not suppress luciferase or Renilla activity per se (Flentke et al. 2011).
The loss of β-catenin/TCF transcriptional activity was mediated by ethanol's calcium transient. Bapta-AM treatment prior to ethanol challenge rescued β-catenin/TCF transcriptional activity and prevented the ethanol-mediated transcriptional loss (102 ± 19%, p < 0.05 vs. ethanol-only, Fig. 2e). The calmodulin inhibitor calmidizolium similarly prevented the loss of β-catenin/TCF transcription following ethanol treatment (146 ± 45%, p < 0.05 vs. ethanol-only, Fig. 2E). Commensurate with their effects on β-catenin content, Bapta-AM and calmidizolium also enhanced β-catenin/TCF transcriptional activity in the saline controls (p < 0.001), suggesting that calcium signals might normally govern β-catenin stability in neural crest progenitors, in addition to mediating the effects of ethanol.
This loss of β-catenin/TCF transcriptional activity was responsible for the neural crest cell death following ethanol exposure. In ethanol-treated embryos, transient transfection of neural crest with ectopic β-catenin substantially reduced the numbers of apoptotic neural crest, identified by their coexpression of LTR+Sox9+, as compared to ethanol-treated controls that were untransfected or were transfected with the eGFP parent construct (4.3 ± 3.4% vs. 24.2 ± 8.3%, p < 0.05 vs. ethanol-only, Fig. 2f). Conversely, over-expression of the TCF construct ΔTCF, which lacks the β-catenin binding site and thus has dominant-negative activity against the β-catenin/TCF transcriptional complex, caused substantial neural crest losses following both ethanol (25.7 ± 5.6%) and saline treatment (24.8 ± 5.3%, for both p < 0.05 vs. saline-only). The effects of ΔTCF and ethanol were not additive, suggesting their actions were convergent and not separate. Transfection with the parent eGFP construct did not alter the effects of ethanol or saline on neural crest survival.
CaMKII destabilizes β-catenin in ethanol-treated neural crest
In other model systems, several calcium-dependent signals have been shown to alter β-catenin expression, protein stability, and transcriptional activity, including CaMKII, Protein Kinase C, calpains, and JNK. GSK3β is also a potent negative effector, although its activity is not necessarily linked to calcium signaling. Several of these [(GSK3β, CaMKII, Protein Kinase C (PKC))] are ethanol responsive (Lee and Messing 2008; French and Heberlein 2009; Liu et al. 2009; Garic et al. 2011). We investigated which of these effectors might contribute to the calcium-mediated reductions in β-catenin protein and cell survival in these ethanol-treated cells. To narrow the list of candidates, we performed a screen of small molecule inhibitors that were selective for these effectors and investigated their ability to prevent the ethanol-induced neural crest death.
We first explored the potential involvement of CaMKII because, in our previous studies, we showed that it is a downstream target of ethanol's calcium transient, and that inhibition of CaMKII activity prevents the ethanol-induced neural crest death studied here (Garic et al. 2011). Consistent with that work, we found that inhibitors selective against CaMKII (myrAIP; Ishida and Fujisawa 1995), or its upstream activator calmodulin (calmidazolium), both prevented the ethanol-induced cell death as shown by the significant reductions in the number of LTR+ cells within rhombomere 4, which normally does not experience significant cell death (p < 0.001 vs. ethanol-only; Fig. 3). This represented an effect upon neural crest because myrAIP significantly enhanced the survival of Sox9+ neural crest following ethanol treatment (Ethanol + myrAIP: 35.6 ± 5.1 Sox9+ cells, Ethanol-only: 22.3 ± 2.6 Sox9+ cells, p < 0.001; Fig. 4a–c, g) and reduced the percentage of Sox9+LTR+ cells within rhombomere 4 (Ethanol + myrAIP: 9.2 ± 5.8%, Ethanol-only: 44.2 ± 6.1%, p < 0.001; Fig. 4h). In contrast, inhibitors directed against PKC (Go-6983, Ro-32-0432), Calpain (ALLN, Inhibitor V), JNK (JNK Inhibitor II), and GSK3β (BIO, SB-415,286) all failed to prevent the pro-apoptotic effects of ethanol. For those inhibitors, both the number of LTR+ cells in rhombomere 4 (Fig. 4) and the absolute number of Sox9+LTR+ neural crest did not differ from ethanol-only treatment (Fig. 5, Fig. 6d–f). These data were consistent with our previous findings that CaMKII activity is an essential component of the ethanol-induced apoptosis in neural crest (Flentke et al. 2011). It also suggested that other effectors of calcium signaling and/or β-catenin stability, including GSK3β, calpain, PKC, and JNK, were unlikely participants in this model of ethanol's neurotoxicity.
We investigated which of these calcium effectors contributed to the β-catenin loss in ethanol-exposed neural crest. Pre-treatment with either calmidazolium or myr-AIP prior to ethanol challenge counteracted ethanol's effects and normalized the β-catenin protein expression within the ethanol-treated neural folds (Fig. 5a–d). The CaMKII and calmodulin inhibitors also restored the transcriptional activity of β-catenin/TCF within ethanol-treated cells to the level observed in controls, as evaluated using the pBARL TopFlash reporter construct (calmidazolium + ethanol: 184 ± 51% p < 0.05, myr-AIP + ethanol: 98 ± 23% p < 0.001 vs. ethanol-only: 43 ± 18%, Fig. 5g). In contrast, inhibitors selective for GSK3β (SB415,286; Fig. 5f) or calpain (ALLN; Fig. 5e), although known to modulate β-catenin protein stability, failed to override ethanol's effects upon β-catenin and the protein levels remained low within the ethanol-treated neural fold. These findings, combined with our previous demonstration that β-catenin loss was responsible for the ethanol-mediated neural crest death, indicated that ethanol's activation of CaMKII was responsible for the β-catenin loss within this cell population.
CaMKII can phosphorylate β-catenin
Little is known about how CaMKII might destabilize β-catenin. Ethanol treatment does not alter β-catenin transcript levels in these cells (Flentke et al. 2011). Moreover, β-catenin protein disappears from these cells within 2 h of ethanol addition (Flentke et al. 2011), implicating a post-translational effect. Direct phosphorylation of β-catenin by the protein kinase GSK3β targets β-catenin for proteosomal transport and degradation. We considered whether CaMKII might possess similar kinase activity against β-catenin; such activity has not been reported previously. In silico analysis of the amino acid sequences of chick and human β-catenin revealed four conserved sites that contain the CaMKII phosphorylation motif RXX(T/S). These were located at Threonine 332, Serine 389, Threonine 472, and Serine 552 (Fig. 6a). An additional fifth site at Serine 96 of chick β-catenin was not present in human. β-catenin from both species also shared four conserved sites containing the putative CaMKII binding motif LXXLL, and these were located starting at Leucines 156, 405, 536, and 640. Incubation of purified human β-catenin with activated CaMKII and [γ32P]-labeled ATP, followed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis, revealed the presence of a 85 kDa phosphorylated protein that was absent from samples incubated with inactive CaMKII (Fig. 6b). As a positive control, the protein D52, a thrombin-cleaved, known CaMKII target, produced a 32P-labeled protein when incubated with activated CaMKII but not with unactivated CaMKII. This result indicated that CaMKII was capable of directly phosphorylating β-catenin.
To establish which of the putative phosphorylation sites in β-catenin could be phosphorylated by CaMKII, we individually mutated the four target sites (T332A, S389A, T472A, S552A) in human β-catenin and tested the ability of activated CaMKII to phosphorylate each mutant protein. We found that β-catenin proteins that harbored the individually mutated sites could still be phosphorylated by CaMKII. Importantly, sensitivity to phosphorylation was not equivalent among the four mutants. The lowest phosphorylation with 32P was observed with the β-catenin S552A mutant, and moderate 32P labeling was observed with the T332A and T472A β-catenin constructs (Fig. 6c). The S389A mutant had robust phosphorylation labeling that did not appreciably differ from the results obtained with equimolar wild-type β-catenin. These results suggested that CaMKII may have targeted more than one site within β-catenin, and phosphorylation at T332 and T472 might depend on the phosphorylation status at S552. It additionally suggested that S389 was not a phosphorylation target of CaMKII. We then tested the ability of activated CaMKII to phosphorylate the β-catenin triple mutant T332A/T472A/S552A. As compared with equimolar wild-type β-catenin, mutation at T332A/T472A/S552A substantially abolished the ability of CaMKII to phosphorylate β-catenin (Fig. 6d), demonstrating that these sites represent bona fide targets of CaMKII's protein kinase activity. Because a faint band was still observed in the triple mutant protein, it further suggested there might be an additional, cryptic phosphorylation site within the protein.
β-Catenin is a central regulator of proliferation, differentiation, and migration in many cell populations, including those of neural origin. A major finding of this study is the demonstration that CaMKII can directly phosphorylate β-catenin, and this represents a novel mechanism by which calcium and its downstream kinase effectors could govern canonical Wnt signaling and thus control cellular fate. Ethanol's dysregulation of β-catenin via CaMKII also represents a novel mechanism by which ethanol affects cellular activity and survival. In addition, these findings highlight β-catenin and CaMKII as central mediators of ethanol's neurotoxicity within early neural progenitors. We and others showed previously that ethanol causes significant apoptosis within neural crest progenitors and these cellular losses contribute to the neurochristopathies that partly typify individuals who experienced PAE during early development. Our past work has separately shown that two upstream events that are necessary for this apoptosis are the calcium-mediated loss of transcriptionally active β-catenin (Flentke et al. 2011) and the activation of CaMKII (Garic et al. 2011). Here, we show that these events are mechanistically linked and that CaMKII is responsible for the loss of transcriptionally active β-catenin and subsequent cell death. The ability of CaMKII to directly phosphorylate and thereby modulate β-catenin's transcriptional activity, which is essential for neural crest survival, suggests a mechanism by which these cells are sensitive to ethanol during this developmental period.
The Wnt/β-catenin signaling cascade is a crucial regulator of cellular activities within neural progenitors, including brain patterning, neurogenesis, plasticity, and stem cell production, and disruptions in those signals contribute to disorders also seen in FASD including microcephaly, learning impairments, and autism (De Ferrari and Moon 2006; Toledo et al. 2008; MacDonald et al. 2009). Growing evidence indicates that altered canonical Wnt/β-catenin signaling contributes to ethanol's damage as revealed by expression screens of ethanol-treated cell populations (Himes et al. 2008; Chen et al. 2010; Serrano et al. 2010). This broad dysregulation within disparate tissues impairs processes including bone healing (Chen et al. 2010; Lauing et al. 2012), neural differentiation (Vangipuram and Lyman 2012), and cell survival (Yeh et al. 2008; Flentke et al. 2011), and suggests that β-catenin/Wnt dysregulation may be a fundamental mechanism underlying ethanol's pathogenicity. Further supporting this are recent demonstrations that Wnt/β-catenin protects hepatocytes against apoptosis and pathogenesis in alcoholic liver disease (Liu et al. 2012; Tao et al. 2013). While the β-catenin losses in this acute exposure model are clearly mechanistic, reports of increased β-catenin in chronic exposures could reflect compensatory responses to these earlier β-catenin losses. Further investigation into this understudied ethanol target will provide additional insights into how its dysregulation contributes to ethanol's pathogenicity.
This work is also the first demonstration that CaMKII is capable of directly phosphorylating β-catenin. Given that β-catenin contains multiple, conserved CaMKII recognition sequences, and that small molecule CaMKII inhibitors selectively stabilize cellular β-catenin protein and its transcriptional activity, these findings suggest their interaction is functionally relevant. In the non-canonical Ca/Wnt pathway, ligands such as Wnt5a interact with Frizzled receptors, which act, in this context, as G-protein coupled receptors to elevate intracellular calcium and antagonize canonical Wnt/β-catenin signaling (Kohn and Moon 2005). The CaMKII/β-catenin interaction may represent a novel mechanism by which calcium signals, perhaps including those originating from non-canonical Ca/Wnt signaling, control cellular function. CaMKII negatively regulates β-catenin indirectly, through its phosphorylation and stimulation of the TAK1-NLK mitogen-activated protein kinases, which in turn phosphorylate TCFs to inhibit β-catenin/TCF-mediated transcription (Ishitani et al. 2003). β-Catenin contains four CaMKII binding motifs (LXXLL) and four putative phosphorylation sites (RXXT/S) that are 100% conserved between chick and human, despite having diverged ~ 300 million years ago. All four predicted sites reside within the Arm repeat structure and three of these (T332, T472, S552) could be phosphorylated by CaMKII, suggesting they are functional. The S552 site was previously identified as a phosphorylation target for Akt and PKA (the latter, weakly); paradoxically, pS552 enhanced nuclear β-catenin accumulation and β-catenin/TCF/LEF transcriptional activity (Fang et al. 2007; Taurin et al. 2006). Given that CaMKII destabilized nuclear β-catenin and reduced its transcriptional activity, it is possible that pS552 may function to integrate the competing activities of CaMKII and Akt and thereby control canonical Wnt sensitivity.
It is unclear why the direct phosphorylation of β-catenin by CaMKII has not been previously identified. One explanation is technical. The β-catenin form often used for protein expression studies is a glutathione-S-transferase (GST)-fusion protein. In pilot studies (Flentke and Smith, unpublished) we found that CaMKII, at best, only weakly phosphorylated the β-catenin-GST fusion construct, perhaps because of steric bulk from the GST, which comprises approximately 30% of the protein's mass. In support of this interpretation, CaMKII readily phosphorylated the β-catenin-HisTag construct that had its 4.4 kDa HisTag removed. The details of how the CaMKII phosphorylation sites regulate β-catenin stability, both individually and through potential interactions with each other, are the focus of a separate investigation.
We found no evidence that two additional effectors of β-catenin stability, PKC and GSK3β, contributed to the β-catenin losses and cell death caused by ethanol here. PKC negatively regulates canonical Wnt signaling through phosphorylation of β-catenin at S33, S37, and S45 (Gwak et al. 2006). PKC is also stimulated by ethanol in some neuronal cell lines (Lee and Messing 2008). However, ethanol treatment does not increase PKC activity in this neural crest population, nor does PKC inhibition prevent their apoptosis (Garic et al. 2011), and that PKC inhibition could not stabilize β-catenin here is consistent with that work. Another significant, Wnt effector, GSK3β, is not known to be calcium-dependent, but in certain neuronal populations acute ethanol exposure significantly enhances its activity to mediate neuronal apoptosis (French and Heberlein 2009; Liu et al. 2009). Thus, it was unexpected that it did not contribute to the cell death studied here, nor did its inhibition stabilize β-catenin. However, a similar lack of GSK3β participation occurs in ethanol-exposed neonatal mouse brain, in which lithium but not SB-415,286 blocks neuronal apoptosis independent of GSK3β activation (Zhong et al. 2006; Young et al. 2008). Thus, ethanol's stimulation of neuronal GSK3β may be lineage-dependent. Overall we conclude that ethanol's dysregulation of β-catenin did not involve signaling through the canonical Wnt pathway, an interpretation that is consistent with our prior demonstration of no expression differences in the canonical Wnt pathway (Flentke et al. 2011).
How does the calcium/CaMKII-mediated loss of transcriptional β-catenin result in cell death? β-Catenin is indispensible for neural crest survival and normal craniofacial development through its induction of genes essential for neural crest development including Wnt6 and FoxD3 (Stuhlmiller and Garcia-Castro 2012). We showed previously that ethanol sharply down-regulates these inductive signals (Flentke et al. 2011), and their repression could reduce neural crest progenitor pools and contribute to neural crest deficits in FASD. Ethanol also significantly reduces slug/snail2 (Flentke et al. 2011), an anti-apoptotic transcriptional repressor that controls the expression of caspase and Bcl-2 family members (Tribulo et al. 2004). Later in development, β-catenin participates in neural crest migration (De Calisto et al. 2005) and differentiation into sensory neurons (Lee et al. 2004), suggesting this dysregulation may also contribute to additional neurochristopathies associated with PAE.
In summary, transcriptional β-catenin is an immediate and important target of PAE. The loss of β-catenin/TCF/LEF signaling mediates the apoptosis of ethanol-exposed neural crest, and, given the importance of canonical Wnt signaling to neural crest induction and differentiation, this repression may also play a broader role in FASD neurochristopathies. Because β-catenin also has critical roles in neuronal development, its disruption by ethanol may underlie additional FASD pathologies and, as suggested by recently published studies of liver, in adult alcoholics. It is unlikely here that ethanol targeted canonical Wnt pathways. Rather, ethanol is teratogenic because it activates pathways that mimic or converge upon canonical Wnt and Ca/Wnt pathways within neural crest. These pathways, and how ethanol interacts with them, are shown in Fig. 7. Ethanol's activation of CaMKII, via stimulation of PLC/PIP2-mediated calcium release, mediates the loss of transcriptionally active β-catenin and its trophic support. Given that both acute and chronic ethanol exposures modulate intracellular calcium in diverse cell types, the calcium-mediated destabilization of β-catenin described here might also underlie the seemingly diverse alcoholic pathologies in other fetal and adult tissues.
Supported by NIH MERIT Award R37 AA11085 to SMS. The authors have no conflict of interest to declare.